This is my submission for @swyx's Kill My SaaS Competition!
Live Deploy Preview: http://loop.dharmicdata.org
Repo: https://github.com/ChaiWithJai/open-speaker-operations
Proof of Work (livestream): https://x.com/ChaiWithJai/status/2087647960264003911?s=20
Unique Considerations: Designed using Devin DeepWikis to research what off-the-shelf system of record I could extend. And then used Buzz by @jack to create an API for chatting with the system and providing shareable URLs. The idea was to seperate the "system of record" from the "system of actions" for a SaaS that is meant to be used by both humans and AI agents alike.
Jai Bhagat: Writing Something Worth Reading https://x.com/i/broadcasts/1lKQRWRmWaWGE

out of curiosity, i tasked fable to figure this thing out and it looks like the "saas killer" is done.
▲ it maxed out Fable 5 + used most of Codex limits
▲ most of the work was done within 45h session running on claude code
▲ other sessions lasted for 17, 14 and 10 hours
▲ there were 183 sub agents used in the process
▲ without subs, tokens would cost $4246
codebase scores 100% on all the benchmarks provided by @swyx and (arguably) the ui looks quite good
best part, I did spend maybe 1h total over the last 4 days steering the agents.
on one hand, this was almost too easy. really feel like software is dead
BUT
i really doubt most of the companies would be willing to now take the responsibility for deploying and maintaining a software like this
curious to see what Shawn will do about it
swyx: ## eval competition idea: Help kill my SaaS
my team is proposing to pay >$40k/year for enterprise saas we have never used and will never be able to customize.
as a smol business owner, this feels shitty.
thinking of doing a small remote hackathon:
- i cover $1000 in tokens

Replit made Inc.'s Top 5000 at #24, our first time on the list.
Read the story 👇

Fascinating. Some data centers came to a struggling town in Washington state. Did it run out of water? Was the community destroyed?
Let’s see. The town…built a new high school, hospital, library, sewage systems, and police & fire stations. Poverty fell from 29% to 6%.
Oh.

the knowledge economy is getting bifurcated into two subeconomies:
1. the things that require the current frontier models to do well
2. the things that are indistinguishable in perf between Deepseek v4 pro / Grok 4.6 vs GPT 5.6 sol / Claude Fable
I’d argue almost everything in the existing economy is now in the latter camp.
excitingly, some of the coolest startup ideas are in the former.

Awesome day in terms of new model releases from both Deepseek and Grok. Both of these new model updates represent huge jumps in capability at insanely low costs.
This is truly Jevons paradox for AI playing out. By bringing down the cost of AI, demand will continue to accelerate. When I talk to enterprises, they have vastly more use-cases for what they’d like agents to handle than generally what they’re able to budget for today.
They want agents to scan their codebases for security issues. They want agents to review all their documents. They want agents to process information in a large portion of their workflows. And so on. When we bring down the cost of AI or drive up capability levels, it opens up all new use cases in the enterprises, which drives even greater demand.
This is also great for the applied AI layer. The more choice of models - especially those tuned for different types (or costs) of work - the more value there is in being at the layer that can route and optimize based on the task. Great for the entire stack right now.
Yuchen Jin: 3 frontier models in one day!
- Grok 4.6: Fable 5-level, but 85% cheaper.
- Qwen3.8-Max: 2.4T params, 95B active. Weights are out.
- DeepSeek-V4-Pro-0813: weights could drop any time. Heard it’s good, not just benchmaxxing.
Competition is great for consumers and businesses.

paired @mattshumer_'s gauntlet loop with my @threejs agent skills to build a Quake inspired CTF game in an underwater base
Opus 5 handled the initial build with ultracode and generated assets with @tripoai @ElevenLabs @NanoBanana - my honest take 👇
the game was about 75% functional after the initial run, but required additional loops and refinement to iron out bugs.. it's interesting way of building games, but I prefer building them out slowly piece by piece where you have more design control.. either way a fun experiment to showcase model capabilities!

agent traces are the most important new input to a self-improving company.
but a pile of traces doesn't improve anything on its own. the full loop takes three components:
1. a company brain - shared memory of what the team knows: decisions, context, the why behind the work
2. agent traces - the raw thought process: what every human and agent actually tried, what worked, what got thrown away
3. a learning loop - the missing piece at most companies: turning those traces back into better context, better skills, and better next runs
most teams have some of 1 and are starting to accumulate 2. almost nobody has 3. the companies that close the loop get a little better every day, recursively.

你相信吗，硅谷大佬，也有曾经看走眼的时候。
@ycombinator CEO @garrytan 很坦诚地分享过自己年轻时的一次“大失误”。
刚从 Stanford 毕业、还在微软工作的他，被 Peter Thiel 亲自劝说离职，加入当时刚刚起步的 @PalantirTech 。
Peter 甚至问他：“你在微软一年赚多少？”
他说，$72,000。
Peter 当场开了一张同等金额的支票给他：
“拿着它，辞掉微软。这样你就没有风险了。”
23 岁的 Garry 想了想，还是拒绝了。
因为他觉得 - “我明年可能就升到 Level 60 了。” 😅
后来 Garry 把这个选择称作自己的 $ 2 -4 billion mistake!
当然，故事并没有停在这里。
他后来还是加入了 Palantir，成为第 10 号员工；再后来去了 YC、创办 Initialized Capital，并成为硅谷最成功的早期投资人之一。
Coinbase、Instacart、Flexport、Rippling、Cruise……
都是他很早就看中的公司。
尤其 Coinbase — 他早期投资的 $300,000，2021 年他自己算这笔股份一度价值 $2.4 billion，约 6000 倍回报
所以这个故事让我励志的地方，就在于一个人可以在一次选择上错过几十亿美元，却依然可以在人生后面的选择里，创造更多的价值。
我们常常会对过去某一个决定耿耿于怀：
“如果当时我接受了那个 offer……”
“如果当时我买了那只股票……”
“如果当时我没有离开那个人……”
“如果当时我抓住那个机会……”
但人生不是一道只有一次作答机会的选择题。人生更像是一连串的期权。错过了一次，不代表你错过了整个人生。
真正重要的，不是确保自己永远不犯错，而是犯错以后，依然保持好奇，保持行动，保持判断力，继续让自己出现在新的机会面前。
You can miss Palantir and still find your Coinbase.
人生的容错率，可能比我们想象中高得多。
Carry on. ❤️
a16z: Garry Tan says turning down Palantir for a Microsoft promotion was a $2–4 billion mistake:
"Joe Lonsdale and Stephen Cohen were my fraternity brothers at Stanford... they were interns at Peter Thiel's hedge fund."
"They said, 'Let's go get Garry. He's up in Seattle.' They flew

My @aiDotEngineer talk on lessons learnt from studying memory systems for a year:
> the evolution of chatgpt and claude memory over the years
> their architectural convergence
> why memory cannot be outsourced
> how memory is a function of compute
> continuous learning for individuals
> why we have a massive context problem

my entry into the arena: https://unsession.dev
- full sandbox environment
- hosted on @Cloudflare (easy to self-host)
- full MCP w/ DCR and API docs
...and lots of TLC and polish
ty for the fun challenge @swyx 🙏
swyx: ## eval competition idea: Help kill my SaaS
my team is proposing to pay >$40k/year for enterprise saas we have never used and will never be able to customize.
as a smol business owner, this feels shitty.
thinking of doing a small remote hackathon:
- i cover $1000 in tokens

Re again i used the same thing. i told opus 4.6 to modify @mattshumer_ 's gauntlet prompt
idk but this prompt just works everywhere perfectly
opus 4.6 actually understands it better than 4.7 and 4.8
(don't tell me about 5 btw lol)

As Trump decimates US aid for poor people around the world, he is sending some of the funds to right-wing think tanks in Europe to promote his ideology. https://trib.al/RJQ8RRZ

So @swyx announced his “Kill My SaaS in one weekend” contest, where you could win $10K to replace his $40K/year SaaS. On Saturday morning, I learned that it was to replace their conference CFP software.
OMG. I've run about 24 conferences over 12 years, but something I rarely talk about is how much I've hated the software we've had to use — over the last decade, we've used 5+ CFP tools, which manage the process of taking speaker submissions and running a review process, some MUCH worse than others.
With each, we've had to build so many workarounds, things like Basecamp, Trello, Google Sheets, Zapier, and I've written entire new apps to be the reviewer frontend. (BusyConf was the one we loved, but they went out of business.)
So I jumped into the contest (Margueritte Kim and I will donate it all to a STEM charity if we win), and less than 24 hours later, I built CurtainCall CFP.
This has been the craziest dev experience of my career — and when Swyx released his eval harness, the entire project became a hill-climbing exercise, and began one of the craziest infrastructure experiences of my career. (Will write more about that later!)
But it's in production! The Enterprise AI Summit is in Charlotte on Oct 7–8, and there are a few slots that may become available! If you want to submit an experience report talk, submit your proposal!
(And you can see the 14 amazing announced speakers (more coming) here, too!)
https://curtaincallcfp.com/program/enterprise-ai-summit-charlotte-2026
swyx: ## eval competition idea: Help kill my SaaS
my team is proposing to pay >$40k/year for enterprise saas we have never used and will never be able to customize.
as a smol business owner, this feels shitty.
thinking of doing a small remote hackathon:
- i cover $1000 in tokens

JEPA is one of the hottest buzzwords in machine learning right now.
Are you still unsure what that means for video models?
Then I've got you covered!😌
In this week's post, we examine the shift in video self-supervised learning from pixel-space to latent-space modeling.

My submission for @swyx Kill my SaaS!
This was fun, and definitely different than what I normally build nowadays.
Given more time I would definitely polish design/UI, but I’m quite proud of how everything works under the hood, and what I was able to accomplish with the time constraint!
Check it out: https://github.com/maddiedreese/ProgramLoom and
https://ProgramLoom.com
Fully free and open source!
swyx: ## eval competition idea: Help kill my SaaS
my team is proposing to pay >$40k/year for enterprise saas we have never used and will never be able to customize.
as a smol business owner, this feels shitty.
thinking of doing a small remote hackathon:
- i cover $1000 in tokens

This was a fun hackathon put on by @swyx .
What happens if you put out a 10K bounty to vibe code a narrow vertical b2b software product (in this case conference organizer software) by cloning the dominant software?
The quality of the entries is incredible. I'm imagining that the amount of open-source software produced from this event could be incredibly useful for the entire conference industry going forward. Boom: industry transformation for $10,000.
Could this be repeated ad infinity to all the software verticals that used to be too narrow to get much attention or competition?

really enjoyed building this, just for the heck of it!
my workflow:
1. parse Google doc from @swyx (incl description + screenshots)
2. perform reconnaissance on the target SaaS (full crawl of their site + learning pages via Firecrawl)
3. generate plan using Codex GPT 5.6 Sol High
4. implement using Opus 5 via Cursor (GRIND mode engaged)
5. run for 4.5 hours straight❗️
6. iterate + feedback via Cursor CLI (~10 turns)
judge the result yourself
(link in the tweet below)
swyx: ## eval competition idea: Help kill my SaaS
my team is proposing to pay >$40k/year for enterprise saas we have never used and will never be able to customize.
as a smol business owner, this feels shitty.
thinking of doing a small remote hackathon:
- i cover $1000 in tokens

A lesson I just taught a Flexport employee is that there are two lenses you can look at the world through:
1) Doing more work is bad
2) Solving problems is how you make money
Always and forever use lens 2. More work is good if it lets you solve important problems.

The OpenAI Foundation is launching AI for Civil Society and Philanthropy to help nonprofits, community organizations, and other trusted partners put advanced AI to work on urgent challenges, develop practical solutions and scale what works.
The first effort is a $100 million commitment with Common Health Coalition to launch Breakthroughs to Follow-through to help care teams use AI to reach more patients with proven treatments - starting with the goal of doubling hepatitis C cure rates in Alabama, Illinois, Louisiana, and Massachusetts. Breakthroughs to Follow-through will work to identify patients who have fallen out of care, synthesize records, track progress, and focus limited staff time where it can have the greatest impact.
The OpenAI Foundation wants to empower those who are closest to issues and to put AI to work where it can make a meaningful difference in people’s lives.
https://openaifoundation.org/news/civil-society-and-philanthropy

Just submitted an entry into @swyx's "Kill my SaaS" competition: http://speakerweave.com
It's truly amazing how low the moat is for niche SaaS. In total, feature parity only took ~1.5 days to hit 100% on the eval. Then I spent 2 more days for polish and adding agentic features, a CLI+MCP, and of course integration to @tryeveryai
This was also done with zero domain knowledge of conference software. I'm not a SaaS doomer, and many incumbents have durable moats wrt distribution/brand and handling high-risk things like money, but damn, there are certainly going lots of zeros, especially in the feature-shallow niches.
swyx: ## eval competition idea: Help kill my SaaS
my team is proposing to pay >$40k/year for enterprise saas we have never used and will never be able to customize.
as a smol business owner, this feels shitty.
thinking of doing a small remote hackathon:
- i cover $1000 in tokens

This was one of the best interviews of the year. Takeaways I shared with my team from @davidsenra x @lulumeservey:
- Simplify your message so it spreads.
- Having a strong, compelling mission is key to "going direct." You are building a movement, and you need to believe in yourself and your mission first before others do. Going direct allows you to communicate more clearly than filtering through layers of journalists, editors, etc.
- When you're talking about your company, focus on the human story and how people will benefit; less on the product, more on the benefits it will have for people.
- Fake engagement with social media is worse than low engagement because people see it and think everything you're doing is fake. We are inundated with fake BS, and it's even more important now to stand out.
- Lean into what makes you distinctive. There's a shortage of originality in the world. Don't use templates. Don't take shortcuts.
David Senra: Lulu (@lulumeservey) is the best in the world at what she does.
She’s the go-to comms strategist for Silicon Valley and many of the top founders in tech trust her.
She’s an advocate for going direct, has built a singular career and shares everything she’s learned.
0:00 From

TODCO, an SF affordable-housing nonprofit, watched 15 tenants die of overdoses in infested, unheated buildings while its lobbying budget grew. It also delayed a 495-unit project for two years using a procedural loophole.
It's not an aberration. A small network of nonprofits builds, manages, and allocates the city's subsidized housing, and helps write the rules that keep it scarce.

Workers' share of GDP is at its lowest point on record whereas the corporate share is near record highs.
This dynamic may exacerbate as Trump's fiscal policies disproportionately benefit corporations and the wealthy.

If everyone's building the same thing, we build what's next.
Hear from the team that built Replit Design, and learn why we set out to reinvent the design process for the age of AI.
The next era of design, for everyone.
https://replit.com/design

just passed 100 billion
Will Bryk: Exa is now one of the largest indexes in the world.
We serve 80B pages, track 1.4T urls, and are on track to be Google-scale in early 2027.
It's difficult to know the size of indexes, but we estimate Google to be ~1T, Bing to be ~500B, and Yandex ~200B. Brave mentioned in

Honored to be mentioned alongside @amasad and @HayaOdeh in this @Inc article about Replit.
I owe a lot to @Replit. WhisperAI was entirely built with Replit Agent, allowing me to go from idea to 7-figure ARR in 6 months and build a product now used 24/7 around the world.
Congrats to Replit and its team on ranking #24 on the Inc. 5000. Hopefully WhisperAI will be joining them on the list next year!
Replit ⠕: Replit made Inc.'s Top 5000 at #24, our first time on the list.
Read the story 👇

I still remember the talk @paulg gave to our YC batch in W23.
Here’s what he said:
1️⃣ To find a great startup idea, turn off the three ‘bad idea filters’:
By turning off the ‘bad idea filters’ in your head, it instantly becomes easier for good ones to emerge.
1/ The sanity filter: Often, the greatest startups come out of ideas that originally seemed insane. So don’t shut down ideas because they seem impractical. For eg, when Airbnb started, no one ever thought people would let strangers sleep in their homes.
2/ The money filter: Money is important for a startup, but if you’re able to drive strong growth without generating revenue, don’t discard the idea. For eg: JustinTV (now Twitch), began with a guy live-streaming his life with a camera strapped to his head. There was no revenue stream initially.
3/ The schlepp filter: Don’t dismiss an idea just because it seems like too much work. For eg: Stripe had to handle unholy amounts of legal and regulatory work to get banks onboard their payments platform. Still, they did the hard work, and Stripe solved a problem that hundreds of their contemporaries were afraid to work on. Often, the impractical hard work becomes the moat that protects your startup from competition.
2️⃣ Don’t build for boomers

Try to build something that gets popular amongst the younger generation. If you’re a college student and you find a lot of your peers using what you’re building (and not just because they’re your friends), it could well be the start of something big.
3️⃣ Stop thinking all the best ideas are taken
As founders, we often come up with an idea only to realize another company has already built it. We then find ourselves wishing that such other company didn't exist. You need to throw that mentality out the door. One quote I’ll always take away from PG’s talk is that there are billions of years of existence left. It would be practically impossible (and quite frankly depressing) for all the good ideas to have already run out.
Paul Graham: This is the 47th YC batch. I missed talking to a couple during covid, but except for that I think I've talked to all of them. Strangely enough I actually write new talk notes each time, though I bet there's a huge amount of overlap with what I've said in past years.

YC doesn't have a word for this, but this is what it looks like when I talk to a whole bunch of startups at once. I only talk to each co for about 5 minutes, but it's like office hours: we go right into what they're working on, and often come up with new ideas for what to do.

The case for techno-optimism. Most new technologies feel scary at first. But they have unlocked a standard of living that our ancestors could only dream about.
"All the items in this room were once out of reach; some not yet invented, others too rare or costly for the vast majority of people. Today, most of us lucky enough to live with them walk past without a second thought.
It is to humanity’s credit that we remain restless in the midst of all of this progress. We continue to look forward, pushing the frontier further with new treatments, new tools, and new institutions that will help future generations in ways we can’t even picture yet.
But our lives today are a gallery of past generations’ heroic efforts to do the same. It serves us, and honors them, to recapture whenever possible the old sense of awe at these wonders that have long since become commonplace."
Sam Rose: This might be the most beautiful thing I've read in my life. What a privilege to be alive in 2026.
https://ordinaryabundance.com